Darwin Family¶
The CMA-ES + paper-exact merge methodology. See the
evolutionary-model-mergingrepo and the 5-stage pipeline blog post.
Definition¶
Darwin Family is the training-free weight-space recombination methodology (per arXiv:2605.14386, Kim et al., 2026). It takes two pretrained LLMs and finds, with zero gradient updates, a linear combination of their weights that outperforms either parent.
The mechanism is: - A 14-dim genome (the search vector) - MRI-Trust Fusion (the linear combination formula) - Architecture Mapper (handles cross-arch tensors, skips if mismatched) - CMA-ES (the optimizer for the genome)
The 14-dim genome¶
{
"gamma": 0.5, # MRI-Trust attention weight
"alpha_attn": 0.5, # attention vs FFN balance
"alpha_ffn": 0.5,
"alpha_emb": 0.5,
"rho_a": 0.5, # rank reduction A
"rho_b": 0.5, # rank reduction B
"r0": 0.5, "r1": 0.5, "r2": 0.5, "r3": 0.5,
"r4": 0.5, "r5": 0.5, # per-tensor ratios
"tau": 0.4, # MRI-vs-genome blend
"lambda_reg": 0.01 # regularization
}
How it's used in the OmniSenter pipeline¶
| Stage | What Darwin does |
|---|---|
| Stage 2 | cma_es_evolution.py searches optimal merge weights across 3 specialized Senter-8B variants |
| Stage 3 | The merged Senter-8B becomes the base for sparse upcycling to the 32B MoE |
| Ohm | The Darwin merge is what makes the Ohm engine cheap (seconds per candidate, not hours) |
| External | continuous_evolution.py in evolutionary-training does the same loop externally |
Why it works¶
The Darwin paper's intuition: high-performing models lie in a low-dimensional manifold of weight space, and CMA-ES can navigate that manifold. The reason it works for LLMs: - Transformer blocks have a regular structure - Attention and FFN layers can be averaged across compatible shapes - The Architecture Mapper handles cross-arch mismatches
See also¶
- Repo:
sovthpaw/evolutionary-model-merging - Blog post:
../../blog/the-5-stage-pipeline.md(Stage 2 specifically) - Blog post:
../../blog/generative-darwin-evolution.md(extending to DiT/audio) - Related: Ohm · Omnimodal Fusion · Senter Ohm
- Paper: arXiv:2605.14386